Graph Self Training

Graph self-training (GST) is a semi-supervised learning technique for graph neural networks (GNNs) that addresses the scarcity of labeled data by iteratively assigning pseudo-labels to unlabeled nodes. Current research focuses on mitigating distribution shifts between labeled and unlabeled data, which can arise from biased pseudo-label selection and negatively impact model generalization. This involves developing algorithms that consider factors like node homophily and employing techniques to ensure consistency between the distributions of labeled and pseudo-labeled nodes. Improved GST methods promise more robust and accurate GNN performance in various applications where labeled data is limited, such as social network analysis and recommendation systems.

Papers

October 12, 2024

BANGS: Game-Theoretic Node Selection for Graph Self-Training
Fangxin Wang, Kay Liu, Sourav Medya, Philip S. Yu
Graph Neural Network Self Supervised Learning Semi Supervised Learning Hair Strand Labeling Technique Graph Self Training

July 25, 2024

HC-GST: Heterophily-aware Distribution Consistency based Graph Self-training
Fali Wang, Tianxiang Zhao, Junjie Xu, Suhang Wang
Homophilous Graph Graph Self Training Distribution Consistency

January 18, 2024

Distribution Consistency based Self-Training for Graph Neural Networks with Sparse Labels
Fali Wang, Tianxiang Zhao, Suhang Wang
Graph Neural Network Self Training Pseudo Labeling Sparse Label Shot Node Classification Unlabeled Node Graph Self Training Distribution Consistency

October 23, 2023

GRENADE: Graph-Centric Language Model for Self-Supervised Representation Learning on Text-Attributed Graphs
Yichuan Li, Kaize Ding, Kyumin Lee
Language Model Self Supervised Representation Learning Graph Contrastive Learning Text Attributed Graph Energetic Material Graph Self Training

August 26, 2022

Toward Robust Graph Semi-Supervised Learning against Extreme Data Scarcity
Kaize Ding, Elnaz Nouri, Guoqing Zheng, Huan Liu, Ryen White
Semi Supervised Learning Semi Supervised Data Scarcity Semi Supervised Node Classification Graph Data Augmentation Limited Labeled Data Graph Self Training

January 27, 2022

Confidence May Cheat: Self-Training on Graph Neural Networks under Distribution Shift
Hongrui Liu, Binbin Hu, Xiao Wang, Chuan Shi, Zhiqiang Zhang, Jun Zhou
Graph Neural Network Loss Function Distribution Shift Self Training Unlabeled Node Graph Self Training